That seems to be what some people at the SMPTE conference are proposing (see related article: ISO Video Proposes SDR+ Production Method). In another spin on the idea, Joe Kane and Florian Friedrich want to see the new SDR standard consist of content distributed in 10 bits, with a DCI-P3 color space using BT.2020 encoding coefficients with a peak mastered luminance of 300 cd/m².
To be clear, this proposal targets the SDR mode in HDR sets – not legacy SDR sets. When SDR content is shown on HDR TVs, it is presented with gamma 2.4, Rec 709 colors, 8-bits and may have the luminance range stretched to a peak of around 300 cd/m². Alternatively, some HDR TVs have an SDR-to-HDR mode that expands the luminance range and color volume to create a kind of HDR image. Some people object to this mode as the TV set maker is making creative decisions – not the content creators. “We’re concerned about SDR image quality in the face of HDR, especially where set manufacturers will feel forced to do something with SDR images,” said Kane.
Kane and Friedrich want to propose a third approach. This new SDR master would be derived from the HDR master, so it would be under creative control. Why do you need a new SDR master for use on an HDR set if an HDR master exists? Presumably it is needed for the situations where a consumer has access to the enhanced SDR version, but not the HDR version.
So why choose the above values for a new SDR production standard? First, Kane recognizes that as the percentage of the screen devoted to full white luminance increases, the overall luminance level decreases. Most of the peak luminance specifications for a 1000 cd/m² HDR TV are set with only 10% of the screen at full white. Once this reaches 100%, the luminance drops, but varies greatly from TV to TV, ranging from 150 to 800 cd/m². HDR sets in the SDR mode can mostly produce 300 cd/m² with 10% or 100% white patches. Therefore, 300 cd/m² seems to be a reasonably obtainable level to set the peak white level.
In addition, many phones, tablets and computer monitors can now support display at 300 cd/m² and P3 color gamut. Friedrich thinks that displaying content mastered at 100 cd/m² and 709 colors does not look good on these displays as both the luminance range and color may be expanded. Conversely, displaying content mastered at 1000 cd/m² may not look as intended either. “We need a solution that is optimized for these screens,” says Friedrich.
A third reason is that some content will just not look that good when remastered at 1000 cd/m² in HDR. “This includes analog tape recordings and some films that used less than stellar film stock or that has not aged well,” claims Friedrich. “These may well look better in SDR/P3 at 300 cd/m² and 10 bits.”
The 10 bits per color should be adopted to minimize any contouring artifacts as the potential visibility of contouring goes up as light output goes up.
The DCI-P3 color gamut is chosen as it is the de facto standard today for color grading of current HDR images. Although DCI-P3 isn’t officially a part of BT.2100 – there are only specifications for 709 and 2020 color, colors are encoded using 2020 coefficients and placed inside the 2020 container.
In order for the HDR TV to recognize this enhanced SDR signal, Kane says no new signaling is needed in the HEVC compression encoding. “What is needed is a smart method for using the information that is already present,” explains Friedrich.
HDR or SDR content encoded with the HEVC compression codec has two metadata structures that carry information helpful for the identification of the incoming content and reconstruction of an optimized image. Supplemental enhancement information (SEI) and video usability information (VUI) provide extra information that can be inserted into the bitstream to enhance the use of the video for a wide variety of purposes including HDR. For example, VUI metadata can include items such as the color primaries, matrix coefficients and transfer characteristics while the SEI metadata includes the reference white luminance, extended range of the white luminance and the nominal white level code value. These structures offer a lot of flexibility to enable the display of this enhanced SDR mode.
One concern Kane has is how to get the TV to display colors in a P3 color gamut. One method is to convey the P3 color information in YCbCr but using 709 coefficients. This would require the TV to be manually switched into the P3 color space to properly display the colors. The other method is to convey the P3 colors in YCbCr using 2020 coefficients and use metadata (perhaps in the SEI or VUI blocks) to automatically set the TV to the P3 color mode. Kane and Friedrich still need to evaluate these options a bit more to develop a full proposal.
Fortunately, it appears that most HDR TVs already allow the color space to be set to P3 and the daylight luminance in SDR mode is already around 300 cd/m², so it may only be a firmware change to create this enhanced SDR mode.
Further, the enhanced SDR approach works well when content is played on projectors. They have a more difficult time reaching the 1,000 cd/m² levels of the PQ system but function well in the gamma display mode. Even single-chip DLP projectors can have a 10 bit P3 capability.
Kane and Friedrich plan to produce some content on a new demo disc to show the creative community and TV makers what such content might look like. These are two well-respected technologists, so this idea is worth evaluating. – CC
